international journal of comparative...

http://cos.sagepub.com

Sociology International Journal of Comparative

DOI: 10.1177/0020715208093082 2008; 49; 369 International Journal of Comparative Sociology

Werfhorst Michelle Jackson, Ruud Luijkx, Reinhard Pollak, Louis-André Vallet and Herman G. van de

Comparative PerspectiveEducational Fields of Study and the Intergenerational Mobility Process in

http://cos.sagepub.com/cgi/content/abstract/49/4-5/369 The online version of this article can be found at:

Published by:

http://www.sagepublications.com

can be found at:International Journal of Comparative Sociology Additional services and information for

http://cos.sagepub.com/cgi/alerts Email Alerts:

http://cos.sagepub.com/subscriptions Subscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.co.uk/journalsPermissions.navPermissions:

http://cos.sagepub.com/cgi/content/refs/49/4-5/369 Citations

at Universiteit van Amsterdam SAGE on June 25, 2010 http://cos.sagepub.comDownloaded from

http://cos.sagepub.com/cgi/alerts

http://cos.sagepub.com/subscriptions

http://www.sagepub.com/journalsReprints.nav

http://www.sagepub.co.uk/journalsPermissions.nav

http://cos.sagepub.com/cgi/content/refs/49/4-5/369


369 International Journal of Comparative Sociology 49(4–5)

Educational Fields of Study and the Intergenerational Mobility Process in Comparative Perspective

Michelle JacksonUniversity of Oxford, UK

Ruud LuijkxTilburg University, The Netherlands

Reinhard PollakWissenschaftszentrum Berlin für Sozialforschung (WZB), Germany

Louis-André Vallet CNRS/CREST, France

Herman G. van de WerfhorstUniversity of Amsterdam, The Netherlands

AbstractThis article examines the importance of educational fi eld of study, in addition to educational level, for explaining intergenerational class mobility in four countries: France, Germany, the UK and the Netherlands. Starting from standard models that only include educational level, we increase the complexity of the educational measure by differentiating between fi elds of study within levels. Contrary to our expectations, including fi eld of study does not substantially reduce the partial effect of class origin on class destination. This seems to be due to the limited association between class origin and fi eld choice, and between fi eld choice and class destination. Implications for stratifi cation and mobility studies are discussed.

Key words: class destination • class origin • education • fi eld of study • intergenerational mobility

INTRODUCTION

In recent years sociological research has provided greater insight into the role of education in processes of social stratifi cation and mobility. While, traditionally, social stratifi cation and mobility research has focused on the study of vertical educational differentiation (higher versus lower levels of schooling), sociologists

International Journal of Comparative SociologyCopyright © 2008 SAGE Publications

http://cos.sagepub.comLos Angeles, London, New Delhi and Singapore

Vol 49(4–5): 369–388DOI: 10.1177/0020715208093082




are recognizing the need to move beyond this narrow conception of educational attainment. One extension has been the emphasis on distinctions between vo-cational and generic types of schooling within the same vertical level, as was extensively studied in the CASMIN project (e.g. Brauns et al., 1999; Shavit and Müller, 1998). This distinction has proved to be very useful, and has great rele-vance for studies of social inequalities in educational attainment and the role of education in occupational attainment. Furthermore, the distinction is particu-larly relevant from a cross-national perspective, as our understanding of institu-tional differences in the organization of educational systems is greatly enhanced by this framework.

One area where education plays an important role is in the process of inter-generational social mobility. In trying to explain the association between class origin and class destination, traditional mobility studies have focused primarily on the role of educational attainment. ‘Educational attainment’ is usually oper-ationalized through measures of levels of schooling, sometimes with the inclusion of additional distinctions between general and vocational tracks. In this article we will examine how far employing a yet further elaborated measure of educa-tional attainment improves our understanding of processes of intergenerational social mobility. In contrast to most existing studies (e.g. Breen, 2004; Ishida et al., 1995; Marshall et al., 1997), we move from the standard operationalization of educational attainment in purely hierarchical terms to one that takes both vertical and horizontal differentiation into account. Specifi cally, we distinguish between different fi elds of study within educational levels.

The relevance of educational fi elds of study for issues of social inequality has been demonstrated in a number of research papers. For example, researchers have studied the impact of gender and social origin on choices of subjects (e.g. Bradley, 2000; Dryler, 1998; Hansen, 1997; Jacobs, 1995; Jonsson, 1999; Polachek, 1978; Smyth, 1999; van de Werfhorst et al., 2001, 2003), the impact of fi elds of study on labour market opportunities (Daymont and Andrisani, 1984; Hansen, 2001; Kalmijn and van der Lippe, 1997; van de Werfhorst, 2002a) and on value orientations and lifestyle (Nilsson and Ekehammar, 1986; van de Werfhorst and Kraaykamp, 2001). We therefore have good reason to ask whether fi eld of study is also relevant to the social mobility process.

Looking at education in a more fi ne-grained way, that is, by examining variations in levels and fi elds of education, is important for two reasons. First, it may shed new light on the extent to which advantage is passed on from parents to children through education, or to what extent parents affect their offspring’s opportunities independent of schooling. One central fi nding of social mobility studies is that social class origin affects children’s social class position even after controlling for education. This means that children of advantaged backgrounds have better op-portunities than children of less advantaged backgrounds partly because of their higher level of educational attainment, but also when they have similar levels of schooling. Thus, social origin has both a direct and, via schooling, an indirect effect



Jackson et al. Educational Fields of Study and the Intergenerational Mobility Process 371

on children’s occupational and class attainment (Blau and Duncan, 1967; Breen, 2004; Marshall et al., 1997). But our estimates of the direct and indirect effects of class origin may be misleading if we ignore horizontal educational differentiation. A fuller investigation of educational choices may reveal that a larger share of par-ents’ infl uence is captured by differential educational choices, thereby diminishing the direct effect on children’s opportunities (Erikson and Jonsson, 1998). Second, including horizontal educational differentiations in mobility studies is important because we might gain further insight into the educational choices that children make as part of their mobility strategies (van de Werfhorst, 2002b). In this regard it is relevant to examine whether in some fi elds of study children are more advan-taged by their class background than in other fi elds (Hansen, 1996; Hansen and Mastekaasa, 2006).

Our central research question is, therefore, does including educational fi eld of study in our analyses of the mobility process alter the magnitude of the direct effect of class origin on class destination? The analysis is carried out for four Western European countries: France, Germany, the UK and the Netherlands. This allows us to examine cross-national variations in patterns of educational choice and social mobility.

RECENT RESEARCH ON THE ROLE OF EDUCATION IN THE INTERGENERATIONAL MOBILITY PROCESS

Following pioneering work by Duncan and Hodge (1963) as well as Blau and Duncan (1967), the status attainment literature has for a long time investigated the role of education in the intergenerational mobility process. Figure 1 shows the hypothesized relationship between social origin, educational attainment and social destination under the status attainment model. We see that the mobility process can usefully be conceived of as a set of two pathways. First, fathers (or parents) transmit their occupational status to their children via education: this is so because social background infl uences educational attainment (arrow a), and educational attainment to a signifi cant extent infl uences occupational outcomes in modern societies (arrow b). Second, fathers (or parents) can transmit their occupational status to their children in a direct way, for instance when a transfer of proprietorship occurs. More generally, arrow c represents all the channels by which fathers (or parents) can infl uence the occupational status of their off-spring controlling for education.

In a comparative study on Social Mobility in Europe, Breen and colleagues (2004) tracked trends in social mobility in the last quarter of the 20th century. For six out of 11 countries (France, Germany, the UK, Ireland, Sweden, and the Netherlands), the volume presented detailed studies on the changing relation-ships between class origin, educational attainment, and class destination over time – with the following results (see also Breen and Jonsson, 2007; Breen and Luijkx, 2007).1 First, as regards the path from origins to education (arrow a), class inequality in educational attainment has declined in France, Germany, the




Netherlands and Sweden, but not in the UK and Ireland. Second, the associa-tion between education and class destination, controlling for class origin (arrow b), has declined in all six countries. Third, the association between origin class and destination class, controlling for education (arrow c), has remained constant in Germany, the UK and Ireland, whereas it has declined in the Netherlands. Fourth, there is a compositional effect, in which the general increase in social fl uidity is seen to result from increasing numbers of highly-qualifi ed individuals in the population, for whom the association between origin and destination is weaker. This weaker origin-destination association for highly-qualifi ed individu-als has been demonstrated in a period perspective for the United States (Hout, 1988) and France (Vallet, 2004) and in a cohort perspective for Sweden (Breen and Jonsson, 2007) and Germany (Breen and Luijkx, 2007).

The whole research fi eld reviewed here has measured educational attainment using the CASMIN educational schema that combines level of education with a distinction between general and vocational tracks. It is therefore the primary goal of the present paper to investigate whether introducing the fi eld-of-study dimension sheds new light on analysis of the ‘Origin-Education-Destination triangle’.

FIELDS OF STUDY AND SOCIAL MOBILITY

While the specifi c question of the role of educational fi eld of study in social mobility is rarely addressed, social scientists have long recognized that fi eld of study is important for outcomes of interest to stratifi cation researchers. For ex-ample, fi eld of study has been shown to be important for wage attainment (e.g. Daymont and Andrisani, 1984; Gerber and Schaefer, 2004; Kalmijn and Van der Lippe, 1997; van de Werfhorst, 2002a; Wilson, 1978; Wilson and Smyth-Lovin, 1983). Fields of study vary in the wage returns that they provide, with the profes-sional fi elds (medicine, business, law) usually paying off rather more than fi elds

Origin Destination

Education

b

c

a

Figure 1 Origin, education, and destination: the OED triangle




in social studies and humanities (for more on this, see the contribution of Reimer et al. in this special issue). This pattern has been demonstrated for a number of Western countries.

More recently, sociological studies of inequality and stratifi cation have for-mulated research questions about the impact of social origin on educational fi elds of study, and how this affects intergenerational social mobility (Ayalon and Yogev, 2007; Hansen, 1997; van de Werfhorst et al., 2001). Davies and Guppy (1997) found for the United States that, controlling for academic ability, univer-sity students of lower socio-economic status families more often enrol in ‘lucra-tive’ fi elds of study with a high labour market pay-off. This led them to conclude that, among working-class students who enrol in university, education is seen as a ‘route to upward mobility’. Parents’ cultural resources, operationalized as fam-ily reading behaviour, did not affect enrolment in economically lucrative fi elds, even if no controls were included for academic ability.

Only a few studies have looked at the role of educational fi eld of study in the whole origin-education-destination triangle. Erikson and Jonsson (1998) showed for Sweden that the effect of social class origin on class destination re-duces sharply after controlling for both educational level and fi eld. People with degrees in teaching, social sciences and health had the relatively highest chance of ending up in the service class, while the social sciences also have a relatively high pay off in terms of income.

Van de Werfhorst (2002b) analysed the detailed role of education in the class mobility process, and showed that children of different social classes choose fi elds that improve their chances of attaining the same class as their parents. Children of farmers often choose the agricultural fi eld, children of skilled manual workers often select technical fi elds at the lower and intermediate level, children of self-employed parents often enrol in economically oriented fi elds at the intermedi-ate level, and children of the service class more often enrol in professional fi elds like law and medicine. Additionally, it was shown that a model that included educational fi elds of study (in addition to level of schooling) decreased the par-tial (direct) effect of social origin on class attainment. This is the only study that has thus far examined this relative balance between direct and indirect effects when employing a measure of fi eld of study.

DIFFERENTIAL EFFECTS OF SOCIAL ORIGIN ACROSS FIELDS

Hansen (1996) has studied the differential impact of social origin on wage at-tainment across different educational fi elds and different occupational groups. The general hypothesis underlying this study was that social origin has a dif-ferential effect on wages across subgroups. Because of network effects and dif-ferential cultural traits, people were expected to benefi t most in the fi elds similar to their parents. Some support for the hypothesis was found: for example, those educated in the fi eld of business benefi t most from their fi eld if they originate in




the class of managers and business executives. Along similar lines, Hansen and Mastekaasa (2006) showed that children from advantaged backgrounds perform better in universities than children from less advantaged backgrounds, but par-ticularly in the cultural and professional fi elds of study. This might be explained by the fact that the requirements for performing well in such fi elds are much more vague than in technical fi elds, so that the benefi ts reaped from a cultured background can be more easily put to use. Skills like speaking in public, writ-ing, and argumentation are more important in these fi elds, and people develop these skills partly in their parental home.

Generalizing from this ‘differential advantage hypothesis’ of the Norwegian studies, it can be expected that the impact of social origin on achievement varies across domains. Applying this to social class attainment, it is plausible to believe that the benefi ts of a high social background extend towards the integration into higher-level occupations, but more so in some fi elds than in others. Children ed-ucated in the humanities may benefi t more from a high-class background than children educated in more technically oriented disciplines. This implies more class inequality among the people educated in the humanities in terms of reach-ing attractive social class positions. On the other hand, among people educated in technical fi elds, class attainment is unlikely to be heavily affected by social background. First, because technical fi elds can be seen as ‘routes to upward mo-bility’ (Davies and Guppy, 1997), it is unlikely that cultural capital has achieved a dominant role in selection and allocation processes in technical fi rms. Second, the types of occupations that technical fi elds prepare for demand fewer skills that are partly achieved in the home, such as social and language skills, and personal style. Given that schools are more important for mathematics achieve-ment than for language achievement, whereas families are more important for language achievement (Brandsma and Knuver, 1989; van de Werfhorst et al., 2003), the fi elds that put a stronger emphasis on mathematical skills may be more meritocratic than fi elds where language skills are important.

DESIGN

Data, Variables and Models

We analyse survey data from four countries: France, Germany, the UK and the Netherlands. For France, we use the Formation-Qualifi cation Professionnelle 2003 Survey. For the UK, we use the General Household Surveys of 1991 and 1992. For Germany, we employ the German Socioeconomic Panel (GSOEP) collected in different years. For the Netherlands, we pool several surveys: The Amenities and Services Utilization Survey (AVO) 1999, the Family Surveys of the Dutch Population of 1992, 1998 and 2000, and the Households in the Netherlands survey 1995 (HIN). See Table 1 for the sample sizes.

We distinguish between four educational levels E: primary or lower second-ary, upper secondary, vocational college/short tertiary, and university. The edu-cational classifi cation that we use can be considered to be a simplifi ed version




of the CASMIN educational schema (due to data constraints it was not possible to use the full schema).

For schooling up to lower secondary level it is not possible to distinguish be-tween fi elds of study. However, from upper secondary level upwards we can distinguish between the following fi elds:

• General (only applicable for upper secondary education, as higher qualifi ca-tions are always within a particular fi eld);

• Humanities/Care/Health/Teaching/Social-Cultural;• Technical/Engineering/Sciences/Agriculture;• Economics/Business/Law.

The combined variable on educational level and fi eld of study is called F, and has 11 categories for France and the Netherlands. Because of data restrictions, it has only eight categories for Germany and six categories for the UK.

For origin O and destination D we use the six-class version of the Erikson and Goldthorpe class schema: I–II, IIIab, IVab, IVc, V–VI, VIIab (Erikson and Goldthorpe, 1992). With this version we follow earlier conventions in mobility research. This way, we can strictly build upon common practice, and see to what extent fi eld of study advances our knowledge of the mobility process. Although we acknowledge that a further disaggregation of classes, in particular distin-guishing classes I and II, would be benefi cial, further differentiation was not possible given the sample sizes available to us.

We also distinguish between three birth cohorts C in each country: individuals born before 1951; born in the years 1951–60; born after 1960.

RESULTS

To analyse the association between these categorical variables, we employ log-linear and log-multiplicative models. These models are particularly useful for studying class mobility (e.g. Breen, 2004; Ishida et al., 1995; Erikson and Goldthorpe, 1992; Ganzeboom et al., 1989).

We set up our analyses in two mobility tables. In the fi rst, we have cohort by class origin by educational level by class destination (COED), in the second, we have cohort by class origin by educational level and fi eld of study by class destination (COFD). We fi t a model that allows for cohort changes in the origin – educational variable association and in the educational variable – destination association, but we do not allow for change in the partial origin – destination association, that is,

Table 1 Analytical sample sizes

Netherlands Germany UK France Total

Men 6239 2872 8119 15,620 32,850Women 5889 2332 7173 15,310 30,704

Total 12,128 5204 15,292 30,930 63,554




we estimate {COE CED OD} and {COF CFD OD} respectively. As is standard practice, we fi t models separately for men and women. In most gender-by-country combinations, this model affords the best compromise between parsimony and fi t.2 To get a synthetic view of the strength of the OD association in these models (using effect coding), we follow Hagenaars (1990) who proposes two overall measures: the maximum difference within the corresponding set of log-linear parameters and the arithmetic mean of the absolute parameters (Table 2). While the former measure relies on the extremes and may be sensitive to outliers, the latter one is likely to be less sensitive, as it is affected by all parameters of the OD association.

In each country, the partial OD association is weaker among women than among men, a fi nding that perhaps results from the more frequent use of father’s occupation to measure origin class. Considering the arithmetic mean of the ab-solute parameters, the partial OD association is weakest in the Netherlands and strongest in Germany (for men) or the UK (for women), with France in an inter-mediate position. However, and most strikingly, Table 2 reveals that the strength of the partial OD association is remarkably similar whether we include only level of education, or both level and fi eld of education. Evidently, we do not observe any consistent pattern suggesting that the partial association between origin and destination is signifi cantly reduced when a more detailed measurement of educa-tion is included. The only exception to this pattern is in the case of Dutch men.

In order to test the differential advantage hypothesis derived from Hansen and others (Hansen, 1996; Hansen and Mastekaasa, 2006), we let the partial OD association of the previous model vary across our level and fi elds of education variable, that is, we introduce a three-way interaction between origin, destina-tion, and level and fi elds of education. As the German data are not suffi ciently numerous to carry out these analyses in a reliable way, we now analyse data from the three other countries only. For these three countries, we compare model

Table 2 Overall association measures based on the partial OD log-linear parameters

Level of education only Both level and fi eld of education

Arithmetic Arithmetic mean of the mean of the Maximum absolute Maximum absolute difference parameters difference parameters

Men France 3.846 0.414 3.808 0.406 Germany 4.008 0.597 4.013 0.641 UK 5.302 0.523 5.301 0.524 Netherlands 3.681 0.357 3.448 0.344Women France 2.604 0.318 2.586 0.317 Germany 2.297 0.395 2.343 0.394 UK 3.768 0.438 3.782 0.438 Netherlands 2.271 0.270 2.366 0.276




{COF CFD OD} with model {COF CFD �FOD}. While the former posits that the partial association between origin and destination is constant across cohorts and constant across categories of F (the level and fi elds of study variable) the latter assumes a common odds-ratio pattern, of varying strength across combi-nations of levels and fi elds for the same association (Xie, 1992). We estimate two versions of that log-multiplicative model: an unconstrained one (in which the strength of the partial OD association varies freely over all categories of F) and a constrained one (which assumes that strength varies across fi elds of study, but is constant over levels for the same fi eld).3

In France, the fi t of the unconstrained model is a signifi cant improvement over the constant model for both men (L2 � 26.0, 10 d.f., p � 0.01) and women (L2 � 30.5, 10 d.f., p � 0.001); and the constrained version is not signifi cantly worse than the unconstrained one, for men (L2 � 12.1, 6 d.f., p � 0.05) or for women (L2 � 9.9, 6 d.f., p � 0.10). In the Netherlands, the unconstrained model does not signifi cantly improve over the constant model, for men (L2 � 7.7, 10 d.f., p � 0.10) or for women (L2 � 15.2, 10 d.f., p � 0.10); and the result is the same when the constrained version is compared with the constant model for men (L2 � 2.8, 4 d.f., p � 0.10), but not for women (L2 � 8.1, 4 d.f., p � 0.10). We therefore have clear empirical evidence that the strength of the partial OD association varies across fi elds of study in France, but only limited evidence for the Netherlands. Table 3 presents the corresponding log-multiplicative parameters for both countries.

In France, for both men and women, the constrained model reveals that the strength of the association between origin and destination is weaker in the Humanities/Care/Health/Teaching/Social-Cultural fi eld of education than in the Technical/Engineering/Sciences/Agriculture fi eld. It is also somewhat larger inthe Economics/Business/Law fi eld than in the Technical one. The unconstrained models generally confi rm these results and they also suggest that the difference between the Humanities and Technical fi elds is strongly reduced at the univer-sity level of education. However, the Netherlands displays the opposite pattern: for both men and women, in the constrained as well as the unconstrained model, the partial OD association tends to be consistently stronger among people edu-cated in the humanities than among those who received a technically oriented education, but the overall effect is not signifi cant.

In the UK data, it is only at the university level that fi elds of study can be iden-tifi ed. The fi t of the log-multiplicative model does not signifi cantly improve over the constant model for men (L2 � 7.8, 5 d.f., p � 0.10) or for women (L2 � 7.8, 5 d.f., p � 0.10). Notwithstanding, parameters estimated at the university level suggest a pattern similar to the French one, that is, they are larger in the technical fi eld (0.91 for men and 0.94 for women) than in the humanities fi eld (0.30 for men and 0.92 for women).

All in all, the differential advantage hypothesis receives limited support. Only in the Netherlands is the partial association between class origin and destination stronger in the humanities fi eld than in the technical fi eld, but that difference is




Tab

le 3

Lo

g-m

ultip

licat

ive

par

amet

ers

for

the

OD

ass

ocia

tion

acro

ss le

vels

and

fi el

ds

of e

duc

atio

n (c

onst

rain

ed a

nd

unco

nstr

aine

d m

odel

s)

Fr

ance

T

he N

ethe

rlan

ds

M

en

Wo

men

M

en

Wo

men

Leve

ls a

nd fi

eld

s o

f ed

ucat

ion

Co

nstr

. U

nco

nstr

. C

ons

tr.

Unc

ons

tr.

Co

nstr

. U

nco

nstr

. C

ons

tr.

Unc

ons

tr.

Prim

ary/

Low

er s

econ

dary

1.

00

1.00

1.

00

1.00

1.

00

1.00

1.

00

1.00

Upp

er s

econ

d./G

ener

al

0.76

0.

75

0.51

0.

51

1.19

1.

21

0.77

0.

71U

pper

sec

ond.

/Hum

aniti

es

0.39

0.

33

0.46

0.

69

1.52

1.

31

0.70

0.

80U

pper

sec

ond.

/Tec

hnic

al

0.89

1.

00

0.70

1.

17

1.02

1.

03

0.27

0.

41U

pper

sec

ond.

/Eco

nom

ics

1.07

1.

18

0.74

0.

63

1.12

1.

36

1.02

1.

13S

hort

tert

iary

/Hum

aniti

es

0.39

0.

11

0.46

0.

08

1.52

2.

40

0.70

0.

42S

hort

tert

iary

/Tec

hnic

al

0.89

0.

85

0.70

0.

30

1.02

1.

06

0.27

0.

06S

hort

tert

iary

/Eco

nom

ics

1.07

1.

29

0.74

0.

75

1.12

0.

65

1.02

�

0.29

Uni

vers

ity/H

uman

ities

0.

39

0.52

0.

46

0.62

1.

52

1.37

0.

70

1.04

Uni

vers

ity/T

echn

ical

0.

89

0.54

0.

70

0.64

1.

02

0.85

0.

27

�0.

12U

nive

rsity

/Eco

nom

ics

1.07

0.

77

0.74

0.

96

1.12

1.

65

1.02

2.

04




not consistently signifi cant. Moreover, France and the UK display the opposite pattern, and differences between fi elds of study prove to be statistically signifi -cant in the former country. We will discuss possible explanations of these con-trasting results below.

EXAMINING THE ORIGIN-EDUCATION AND EDUCATION-DESTINATION RELATIONSHIPS

In the previous section, we found that the strength of the partial OD association is much the same whether we use a measure only of level of education, or wheth-er we use a measure of both level and fi eld of study. In order to understand why this should be the case, we now investigate the underlying relationships between origin and education, and education and destination.

In Figure 2, we present the ‘average’ association parameters between social origin and our fi eld of study variable across cohorts.4 The markers in the graph indicate the log-odds of completing a certain fi eld-of-study within a certain level of education compared to having only primary/lower secondary education.5 For example, for Dutch men of class I � II, the odds of holding a university degree in economics are e1.7 � 5.4 times higher than having only primary or lower sec-ondary education, while the odds of having a university degree in the technical fi elds are e1.3 � 3.6 times higher compared to primary/lower secondary educa-tion. Likewise, for Dutch men of unskilled working-class background (VIIab), the odds of graduating from a university in the technical fi elds is only about one sixth (e�1.8 � .17) compared to having only primary/lower secondary education.6

Looking at the coeffi cient patterns, we can easily fi nd support for the fi ndings of previous research on educational inequality. With respect to level of educa-tion, children from the service classes (I � II) always have higher odds of reaching upper secondary or tertiary education than leaving the educational system with compulsory or lower secondary schooling. In contrast, children from the unskilled working classes (VIIab) always have lower odds of holding an advanced degree. For our purposes, it is more interesting to look for systematic patterns in the asso-ciation between origins and fi elds of study within educational levels. The immedi-ate impression is that the coeffi cients for fi elds of study within an educational level are very close to one other: the variations from top to bottom are very limited in the graph. Hence, there seems to be hardly any additional effect of fi elds of study within a given level of education. This fi nding is especially true for France, the UK and for most associations for the Netherlands. Larger differences can only be found in Germany, in a pattern consistent with earlier research: men of lower class origins are over-represented in the technical fi elds (van de Werfhorst, 2002b).7 If we were pressed to search for some regularity in the pattern of association between origins and educational fi eld of study, we might consider the following in relation to the impact of origins on the choice of technical fi elds: while (mainly male) students from classes I � II, IIIab, and IVab tend to be underrepresented in technical fi elds, students with a farm background are more likely to opt for




Figure 2 Log-odds for origin to fi eld of study association

-2-1012

-2-1012

-2-1012

-2-1012

I+II IIIab IVab V+VI VIIab IVc I+II IIIab IVab V+VI VIIab IVc

up

p.s

ec.

sho

rt t

ert.

uni

vers

ity

up

p.s

ec.

sho

rt t

ert.

uni

vers

ity

up

p.s

ec.

sho

rt t

ert.

uni

vers

ity

up

p.s

ec.

sho

rt t

ert.

uni

vers

ity

up

p.s

ec.

sho

rt t

ert.

uni

vers

ity

up

p.s

ec.

sho

rt t

ert.

uni

vers

ity

up

p.s

ec.

sho

rt t

ert.

uni

vers

ity

up

p.s

ec.

sho

rt t

ert.

uni

vers

ity

up

p.s

ec.

sho

rt t

ert.

uni

vers

ity

up

p.s

ec.

sho

rt t

ert.

uni

vers

ity

up

p.s

ec.

sho

rt t

ert.

uni

vers

ity

up

p.s

ec.

sho

rt t

ert.

uni

vers

ity

Netherlands, men Netherlands, women

Germany, men Germany, women

UK, men UK, women

France, men France, women

General Education Humanities Technical fields Economics

technical fi elds (which include the fi eld of agriculture, where many children of farmers end up; van de Werfhorst et al., 2001). And one could observe that children from the petty bourgeoisie seem to choose tertiary-level education in economics more often than children from other classes do. These differences, however, are very small. Hence, the results of this analysis suggest that including a fi eld of study measure in add ition to a level of education measure adds very little to our under-standing of the overall association between social origins and education.

We move on to examine the second association in the OFD triangle that has potential to infl uence the origin-destination association. In Figure 3 we present the ‘average’ log-odds across cohorts for the partial association between our




fi eld of study variable and class destination.8 Controlling for class origin, the log-odds refer to the odds of reaching a certain class position, given a certain educational level and fi eld of study, relative to having primary/lower secondary education. As an example, the odds of Dutch men with a short tertiary degree in the humanities reaching a service-class position (I � II) are about e1.9 � 6.8 times higher than the odds for those with compulsory/lower secondary educa-tion. At the same time, the odds of these men ending up in the unskilled working class (VIIab) are only about one fi fth (e�1.7 � .19) compared to those with only primary/lower secondary education. We can see for all countries that the level of education is important for reaching the higher classes.9 In particular, any form of tertiary education strongly raises the odds of entering the service class (I � II).

Since we are mainly interested in the additional effects of fi elds of study, we need to consider the class coeffi cients within a given level of education across fi elds of study. Upon fi rst glance, the overall picture seems to be somewhat less structured than in the previous graph. However, if we take a closer look at the coeffi cients within educational levels, we fi nd most class coeffi cients to be in a more or less identical order, with the service class (I � II) at the top and the un-skilled working class at the bottom. And, in fact, within a given educational level, most panels show a fairly undiscriminating pattern in the class coeffi cients for the service class (I � II) and for the unskilled working class (VIIab). There is one notable exception to this pattern. At all three levels of education, we observe a generally lower association between technical fi elds and class IIIab (routine non-manuals). This effect is probably due to the composition of the routine non-manual class, which does not include manual employees (manual workers and technicians instead go into the skilled working classes (V � VI)).

Overall, we fi nd that by adding a measurement of fi eld of study to the level of education measure, we do not gain much additional information on the partial as-sociation between education and class outcome. To be sure, the association will be somewhat stronger compared to an association that only considers educational levels. But this effect is quite limited, and probably too small to cause a change in the partial association between origins and destinations in our COFD table.

Taking this fi nding together with our fi nding that fi eld of study adds little to our understanding of the origin-educational attainment relationship (Figure 2), we now have a better appreciation of why we have not been able to fi nd a weak-ening in the partial OD association. Using a combination of level and fi eld of education does not seem to add much information about the underlying partial associations between origin – education and education – destination, at least for the operationalization at hand. We do not fi nd support for the idea that children from certain class origins choose particular fi elds to promote their class attain-ment strategically. And neither do we fi nd much support for the idea that certain fi elds provide a competitive advantage for certain class destinations. It stands to reason that if hardly any more information is provided by the (partial) associa-tions, the partial origin-destination association should not be expected to change to any great extent either.10




Figure 3 Log-odds for fi eld of study to class destination association

-2-10123

-2-10123

-2-10123

-2-10123

upp.sec. short tert. university upp.sec. short tert. university

gen

eral

hum

aniti

es

tec

hnic

al

eco

nom

ics

gen

eral

hum

aniti

es

tec

hnic

al

eco

nom

ics

hum

aniti

es

tec

hnic

al

eco

nom

ics

gen

eral

hum

aniti

es

tec

hnic

al

eco

nom

ics

gen

eral

hum

aniti

es

tec

hnic

al

eco

nom

ics

hum

aniti

es

tec

hnic

al

eco

nom

ics

Netherlands, men Netherlands, women

Germany, men Germany, women

UK, men UK, women

France, men France, women

I+II IIIab IVab V+VI VIIab IVc

CONCLUSIONS

Our aim in this article has been to examine how far introducing a horizontal di-mension (i.e. fi eld of study) to our measurement of education makes a difference when it comes to understanding processes of social mobility. As we argued, the classical approach to studying mobility, as encapsulated in the ‘OED triangle’, implies that the vertical dimension of educational attainment is the only dimen-sion of importance in explaining the association between class origin and class destination. However, in the real world, educational attainment is not a single di-mensional variable, and it encompasses choices made at the horizontal as well as




vertical level; individuals make choices about the fi eld of study that they wish to specialize in, as well as about the level of education that they wish to study for.

Contrary to our expectations, it is not the case that the partial effect of class ori gin on class destination decreases once we control for educational fi eld of study in addition to level of schooling. Only for Dutch men do we fi nd weak evidence in favour of this. Inspecting the underlying patterns of associations between origin and fi eld of study, and between fi eld of study and class destina-tion, we see that these associations do not deliver much more insight into the relationships between origin and education and between education and class destination.

To address the view that class origin may be more benefi cial in cultural and professional fi elds than in other fi elds (what we call the ‘differential advantage hypothesis’, developed from Hansen, 1996; Hansen and Mastekaasa, 2006), we examined whether the impact of class origin on class destination was stronger in the humanities relative to the technical fi eld. We did not fi nd support for this hypothesis. For France and the UK, the results revealed that a stronger impact of class origin is actually found in the technical fi elds, and a weaker effect is found in the humanities. In the Netherlands the pattern confi rmed the differential ad-vantage hypothesis, but the differences across fi elds in the partial OD association were minor. Perhaps the most appropriate conclusion to draw relating to this hypothesis is that the strength of the partial OD association does seem to differ depending on fi eld of study, but that there is no systematic pattern to the strength or direction of this difference across the countries that we have studied.

One important conclusion that our results point towards is that it is unwise to insist that any single measure of a concept will be appropriate in all circum-stances, and for all countries. Clearly, fi eld of study adds to our understanding of the role of education in mobility processes in the Netherlands and France. But it is equally important to recognize that in other country contexts, such as the UK, fi eld of study does not appear to add to our understanding of these processes to any great extent. Therefore, in deciding whether to include horizontal differen-tiation in the measurement of educational attainment, full consideration must be given to the appropriateness of the measure for the country involved (and the question at hand). In some circumstances, this decision may involve a trade-off between full information and parsimony.

So why is it that the inclusion of fi eld of study alongside a vertical measure of educational attainment fails to result in a signifi cant increase in explanatory power in most of the models presented here? One reason might be that there are countervailing forces acting on students (and families) when it comes to the choice of fi eld of study. While some fi elds (such as medicine) have strong links with the labour market, and would therefore appear to be a ‘good bet’ for lower class children who aspire to be successful, these fi elds are often more risky (in terms of high failure rates), more fi nancially demanding, and may be seen as the preserve of the higher classes. Similarly, cultural fi elds may be stereotypically




perceived as those where high levels of cultural capital are important for success, suggesting that the risk of failure might be higher for lower class students, but the admission criteria for these courses is often lower than for courses in tech nical fi elds, which may increase the numbers of lower class students taking courses in cultural fi elds.

To relate this argument to the countries under study in this article, we may note that, given that admissions criteria are more or less the same across fi elds of study in the Netherlands (conditional on the right preparatory school type), it may be that the technical fi elds are relatively more attractive to lower class students than in other countries. This may lead to a greater relevance of the cultural capital argument in the Netherlands. The fact that in the Netherlands the direct impact of class origin on class destination is stronger in the cultural fi elds than in other fi elds, and that this is not the case in France, is in line with this reasoning. In consequence, it is therefore diffi cult to argue that we should expect any consistent pattern cross-nationally, as the way in which educational fi eld of study affects social mobility is likely to depend heavily on the national context of selection into fi elds of study; an issue that merits further research.

A second explanation of our fi ndings might be that our measures of fi eld of study and social class are too aggregated to identify the patterns of interest. Fields of study can only be identifi ed at a relatively aggregated level, yet we know that within these fi elds of study there are important (and possibly consequential) dif-ferences between individual subjects. When it comes to social class, we may well have found a stronger effect of fi eld of study if we had classifi ed occupations using fi ner-grained occupational groups (of the type favoured by Grusky and colleagues, e.g. Grusky and Weeden, 2001; Weeden and Grusky, 2005), rather than the big classes of the EGP schema. The horizontal differentiation by which levels of edu-cation are split up into particular subjects must be seen, in part, as a response to a labour market which demands occupationally specifi c skills and qualifi cations. But big classes encompass a whole range of occupations, so while in any one class some occupations and fi elds of study will have strong links, other occupations within the same class will have weaker links with those fi elds of study. The overall effect of any particular fi eld of study will therefore be an unsatisfactory average effect over the occupations comprising each big class. Subsequently, we would expect that mobility between fi ne-grained occupational groups will be affected to a far greater extent by fi eld of study than mobility between big classes. Data constraints unfor-tunately mean that these more detailed analyses are not feasible, particularly in a comparative context.

A third explanation for our weak effects is that many of the people included in our cross-sectional survey data are educated at (lower) levels where there is no differentiation between fi elds of study. As a consequence of continuing edu-cational expansion in the surveyed countries, in the future we would expect a larger proportion of the population to be educated at higher levels, where fi eld differences may matter more.




To conclude, we have not found evidence to suggest that fi eld of study should routinely be included in cross-national studies of social mobility. However, we would urge caution in using the fi ndings of the analyses presented here to argue that fi eld of study is irrelevant to mobility processes, for the reasons discussed above. The safest conclusion at this time would be to say that the importance of fi eld of study for social mobility is not proven.

ACKNOWLEDGEMENTS

The authors thank Laura Dörfl er for assistance in data preparation, and Richard Breen and Walter Müller for helpful comments on previous versions. During the time this article was being prepared, Michelle Jackson held an ESRC Research Fellowship within the ‘Understanding Population Trends and Processes’ programme. Herman van de Werfhorst is funded by a personal VIDI grant of the Netherlands Organization of Scientifi c Research (N.W.O.).

NOTES

1 Most of the studies in Breen (2004) adopt a period perspective (France, the UK, Ireland, the Netherlands and Sweden), while in the case of Germany, a cohort per-spective is used. Breen and Jonsson (2007) and Breen and Luijkx (2007) use a period by cohort perspective for their studies on Sweden, the UK and Germany.

2 All fi t statistics are available from the authors upon request. 3 For reasons of identifi cation the �-parameter for the fi rst educational category (pri-

mary or lower secondary level) is fi xed at 1. 4 Within the COF table, we fi t {CO CF OF}. The OF parameters are presented in

Figure 2. 5 Note that we use effect coding for class, so the log-odds are indeed log-odds rather

than log-odds ratios. 6 We are not able to show coeffi cients for all four fi elds of study for all levels of educa-

tion. This is partly due to the fact that some fi elds are not offered at some levels of education (for example, there is no general education at the university level in all four countries), and partly due to the sparseness of our data. In order to make sure that our coeffi cients are robust, we only present coeffi cients that are represented by at least fi ve observations per cell. We tested a higher threshold of considering only coeffi cients with at least 15 observations per cell. The overall patterns do not change, and the interpretations based on at least fi ve observations appear to be reliable.

7 The German data set is the smallest data set in our analyses, so the coeffi cients might be volatile. However, if we restrict the results to those cases with more than 15 obser-vations per cell, the pattern for Germany is still more scattered than in other coun-tries. Apparently, social origins have a stronger impact on the choice of fi eld of study in Germany than in France, the UK or the Netherlands.

8 In the COFD table, we model {COF COD FD} and present the FD parameters. 9 As in the previous fi gure, we only display coeffi cients that are based on a minimum

of fi ve observations per cell. 10 We have already ruled out the possibility of strong compositional effects that might

have an effect on the partial OD association.




REFERENCES

Ayalon, H. and Yogev, A. (2005) ‘Field of Study and Students’ Stratifi cation in an Expanded System of Higher Education: The Case of Israel’, European Sociological Review 21: 227–41.

Blau, P.M. and Duncan, O.D. (1967) The American Occupational Structure. New York: Wiley.

Bradley, K. (2000) ‘The Incorporation of Women into Higher Education: Paradoxical Outcomes?’, Sociology of Education 73: 1–18.

Brandsma, H.P. and Knuver, J.W.M. (1989) ‘Effects of School and Classroom Characteristics on Pupil Progress in Language and Arithmetic’, International Journal of Educational Research 13: 777–88.

Brauns, H., Steinmann, S., Kieffer, A. and Marry, C. (1999) ‘Does Education Matter? France and Germany in Comparative Perspective’, European Sociological Review 15: 61–90.

Breen, R.(ed.) (2004) Social Mobility in Europe. Oxford: Oxford University Press.Breen, R. and Jonsson, J.O. (2007) ‘Explaining Change in Social Fluidity: Educational

Equalization and Educational Expansion in Twentieth-Century Sweden’, American Journal of Sociology 112: 1775–810.

Breen, R. and Luijkx, R. (2007) ‘Social Mobility and Education: A Comparative Analysis of Period and Cohort Trends in Britain and Germany’, in S. Scherer, R. Pollak, G. Otte and M. Gangl (eds) From Origin to Destination: Trends and Mechanisms in Social Stratifi cation Research, pp. 102–24. Frankfurt: Campus Verlag.

Davies, S. and Guppy, N. (1997) ‘Fields of Study, College Selectivity, and Student Inequalities in Higher Education’, Social Forces 75: 1417–38.

Daymont, T.N. and Andrisani, P.J. (1984) ‘Job Preferences, College Major, and the Gender Gap in Earnings’, Journal of Human Resources 19: 408–28.

Dryler, H. (1998) ‘Parental Role Models, Gender, and Educational Choice’, British Journal of Sociology 49: 375–98.

Duncan, O.D. and Hodge, R.W. (1963) ‘Education and Occupational Mobility: A Regression Analysis’, American Journal of Sociology 68: 629–44.

Erikson, R. and Goldthorpe, J.H. (1992) The Constant Flux: A Study of Class Mobility in Industrial Societies. Oxford: Clarendon Press.

Erikson, R. and Jonsson, J.O. (1998) ‘Social Origin as an Interest-bearing Asset: Family Background and Labour-market Rewards among Employees in Sweden’, Acta Sociologica 41: 19–36.

Ganzeboom, H.B.G., Luijkx, R. and Treiman, D.J. (1989) ‘Intergenerational Class Mobility in Comparative Perspective’, Research in Social Stratifi cation and Mobility 9: 3–79.

Gerber, T.P. and Schaefer, D.R. (2004) ‘Horizontal Stratifi cation of Higher Education in Russia: Trends, Gender Differences, and Labor Market Outcomes’, Sociology of Education 77: 32–59.

Grusky, D.B. and Weeden, K.A. (2001) ‘Decomposition without Death: A Research Agenda for a New Class Analysis’, Acta Sociologica 44: 203–18.

Hagenaars, J.A. (1990) Categorical Longitudinal Data: Log-linear Panel, Trend, and Cohort Analysis. London: SAGE.

Hansen, M.N. (1996) ‘Earnings in Elite Groups: The Impact of Social Class Origin’, Acta Sociologica 39: 385–408.




Hansen, M.N. (1997) ‘Social and Economic Inequality in the Educational Career: Do the Effects of Social Background Characteristics Decline?’, European Sociological Review 13: 305–21.

Hansen, M.N. (2001) ‘Education and Economic Rewards: Variations by Social-Class Origin and Income Measures’, European Sociological Review 17: 209–31.

Hansen, M.N. and Mastekaasa, A. (2006) ‘Social Origins and Academic Performance at University’, European Sociological Review 22: 277–91.

Hout, M. (1988) ‘More Universalism, Less Structural Mobility: The American Occupational Structure in the 1980s’, American Journal of Sociology 93: 1358–400.

Ishida, H., Müller, W. and Ridge, J.M. (1995) ‘Class Origin, Class Destination, and Education: A Cross-National Study of Ten Industrial Nations’, American Journal of Sociology 101: 145–93.

Jacobs, J.A. (1995) ‘Gender and Academic Specialties: Trends among Recipients of College Degrees in the 1980s’, Sociology of Education 68: 81–98.

Jonsson, J.O. (1999) ‘Explaining Sex Differences in Educational Choice: An Empirical Assessment of a Rational Choice Model’, European Sociological Review 15: 391–404.

Kalmijn, M. and Van der Lippe, T. (1997) ‘Type of Schooling and Sex Differences in Earnings in the Netherlands’, European Sociological Review 13: 1–15.

Marshall, G., Swift, A. and Roberts, S. (1997) Against the Odds? Social Class and Social Justice in Industrial Societies. Oxford: Clarendon Press.

Nilsson, I. and Ekehammar, B. (1986) ‘Sociopolitical Ideology and Field of Study’, Educational Studies 12: 37–46.

Polachek, S.W. (1978) ‘Sex Differences in College Major’, Industrial and Labor Relations Review 31: 498–508.

Shavit, Y. and Müller, W. (1998) From School to Work: A Comparative Study of Educational Qualifi cations and Occupational Destinations. Oxford: Clarendon Press.

Smyth, E. (1999) ‘Educational Inequalities among School Leavers in Ireland 1979–1994’, Economic and Social Review 30: 267–84.

Vallet, L.-A. (2004) ‘Change in Intergenerational Class Mobility in France from the 1970s to the 1990s and Its Explanation: An Analysis Following the CASMIN Approach’, in R. Breen (ed.) Social Mobility in Europe, pp. 115–47. Oxford: Oxford University Press.

van de Werfhorst, H.G. (2002a) ‘Fields of Study, Acquired Skills and the Wage Benefi t from a Matching Job’, Acta Sociologica 45: 287–303.

van de Werfhorst, H.G. (2002b) ‘A Detailed Examination of the Role of Education in Intergenerational Social Class Mobility’, Social Science Information 41: 407–38.

van de Werfhorst, H.G. and Kraaykamp, G. (2001) ‘Four Field-related Educational Resources and their Impact on Labor, Consumption, and Sociopolitical Orientation’, Sociology of Education 74: 296–317.

van de Werfhorst, H.G., De Graaf, N.D. and Kraaykamp, G. (2001) ‘Intergenerational Resemblance in Field of Study in the Netherlands’, European Sociological Review 17: 275–94.

van de Werfhorst, H.G., Sullivan, A. and Cheung, S.Y. (2003) ‘Social Class, Ability and Choice of Subject in Secondary and Tertiary Education in Britain’, British Educational Research Journal 29: 41–62.

Weeden, K.A. and Grusky, D.B. (2005) ‘The Case for a New Class Map’, American Journal of Sociology 111: 141–212.




Wilson, K.L. (1978) ‘Toward an Improved Explanation of Income Attainment: Recalibrating Education and Occupation’, American Journal of Sociology 84: 684–97.

Wilson, K.L. and Smith-Lovin, L. (1983) ‘Scaling the Prestige, Authority, and Income Potential of College Curricula’, Social Science Research 12: 159–86.

Xie, Y. (1992) ‘The Log-Multiplicative Layer Effect Model for Comparing Mobility Tables’, American Sociological Review 57: 380–95.

Michelle Jackson is a Research Fellow of the Centre for Research Methods in the Social Sciences and Nuffi eld College, University of Oxford. Her research focuses on educational inequality, social stratifi cation and social mobility. [email: michelle.jackson@nuffi eld.ox.ac.uk]

Ruud Luijkx is Associate Professor in the Department of Sociology at Tilburg University. He has contributed to international benchmark studies on social mobility and published further in the fi eld of (educational) heterogamy, social inequality, career mobility, and labour market transitions and on log-linear and latent class analysis. He is part of the managerial team of the European Values Study.[email: [email protected]]

Reinhard Pollak is a Research Fellow at the Wissenschaftszentrum Berlin für Sozialforschung (WZB), Germany. His main research areas include social mobility, social stratifi cation, sociology of education, and measures of social inequality. He recently co-edited a book From Origin to Destination: Trends and Mechanisms in Social Mobility and Social Stratifi cation Research (2007). [email: [email protected]]

Louis-André Vallet is a Research Professor in the French National Centre for Scientifi c Research (CNRS) and works within the Quantitative Sociology Laboratory in the Centre for Research in Economics and Statistics (CREST), the research centre of the French Statistical Offi ce. He has been a member of the boards of Revue Française de Sociologie since 1991 and European Sociological Review since 2000. He is mainly interested in social stratifi cation and mobility, sociology of education, and statistical modelling of categorical variables. [email: [email protected]]

Herman G. van de Werfhorst is Full Professor of Sociology at the University of Amsterdam, and is affi liated to the Amsterdam School for Social Science Research (ASSR) and the Amsterdam Institute for Advanced Labour Studies (AIAS). His research focuses on cross-national studies of education and stratifi cation. Together with Robert Erikson and Magnus Nermo from Stockholm University he coordinates a Network of Excellence funded by the European Union, called ‘Economic Change, Quality of Life, and Social Cohesion’ (EQUALSOC).[email: [email protected]].



international journal of comparative...

Documents